Covariation of voice onset time: a universal aspect of phonetic - - PowerPoint PPT Presentation
Covariation of voice onset time: a universal aspect of phonetic realization Eleanor Chodroff 1 , Alessandra Golden 2 , and Colin Wilson 2 1 Northwestern University, Department of Linguistics 2 Johns Hopkins University, Department of Cognitive
1Northwestern University, Department of Linguistics 2Johns Hopkins University, Department of Cognitive Science
92nd Annual Meeting of the LSA Salt Lake City, UT January 4, 2018
Extensive cross-linguistic variation in the realization of speech sounds
e.g., Disner 1978, Lindau 1978, Manuel 1990
e.g., Gordon 2002
e.g., Whalen and Levitt, 1995
e.g., Maddieson 1997, Cho & Ladefoged 1999
Cho & Ladefoged 1999 Keating 1985, 1990, Cohn 1993, Cho & Ladefoged 1999
DORSAL + spread glottis
…
closed, velar spread TONGUE BODY GLOTTIS articulation acoustics Phonetic implementation
[th] mean VOT range: 50 to 150 ms [kh] mean VOT range: 73 to 154 ms [ph] mean VOT range: 63 to 83 ms
Keating 1985, 1990, Cho & Ladefoged 1999
What is the relational structure of cross-linguistic phonetic variation?
closed, velar spread Tongue body Glottis DORSAL – continuant … + spread glottis
[kʰ]
closed, alveolar spread Glottis CORONAL – continuant … + spread glottis
[tʰ]
Tongue tip closed, labial spread Glottis LABIAL – continuant … + spread glottis
[pʰ]
Lips
1) Do the VOTs of [ph], [th], and [kh] vary independently of one another?
VOT
place of articulation [pʰ] [tʰ] [kʰ] VOT place of articulation [pʰ] [tʰ] [kʰ]
VOT
place of articulation [pʰ] [tʰ] [kʰ]
Language 1 Language 2 Language 3
2) Is there consistency in the ordinal ranking of [ph], [th], and [kh]?
VOT
place of articulation [pʰ] [tʰ] [kʰ] VOT place of articulation [pʰ] [tʰ] [kʰ]
VOT
place of articulation [pʰ] [tʰ] [kʰ]
Language 1 Language 2 Language 3
e.g., Maddieson 1997, Cho & Ladefoged, 1999 Variable ranking of [th]: Suomi 1980, Docherty 1992, Whalen et al. 2007, Yao 2009, Chodroff & Wilson 2017
3) Is there a consistent linear relationship among [ph], [th], and [kh]?
VOT place of articulation
[pʰ] [tʰ] [kʰ]
Chodroff & Wilson 2017
Large collection of previously reported stop VOT values Examine relational structure of VOT among stops that have the same laryngeal feature specification*
* not just [+spread glottis], but also [-spread glottis], [-voice], [+voice], etc.
Examined ~350 theses, articles, grammars, and manuscripts Collected stop VOT values from 164 sources 113 languages (149 dialects) 36 language families
Removed:
Removed:
1671 VOT values remained for analysis
Averaged VOT data points with shared place and voice within each study, resulting in 1079 data points
Language Family Languages Data points Indo-European Afrikaans, Armenian (Eastern), Assamese, Bengali, Catalan, Croatian, Danish, Dutch, English, French, Gaelic (Scots), German, Greek (Modern), Hindi, Icelandic, Italian, Kurmanji, Marathi, Nepali, Norwegian, Pahari, Panjabi, Pashto, Persian, Polish, Portuguese (Brazilian), Portuguese (European), Russian, Serbian, Sindhi, Spanish, Swedish, Welsh 557 Sino-Tibetan Bunun, Burmese, Cantonese, Fukienese, Galo, Hakha Lai, Hakka, Hokkien, Karen (Sgaw), Khonoma Angami, Kurtop, Mandarin, Stau, Taiwanese, Wu (Shanghainese) 106 Afro-Asiatic Amharic, Arabic, Dahalo, Hebrew (Modern), Musey 41 Austronesian Belep, Madurese, Malay, Tsou, Yapese 31 Niger-Congo Bowiri, Igbo, Shekgalagari, Swati, Tswana, Zulu 39 Uralic Finnish, Hungarian 21 Na-Dene Apache (Western), Hupa, Navajo, Tlingit 19
Language Family Languages Data points Korean Korean 18 Tai-Kadai Tai Khamti, Thai 18 Tupian Arara, Munduruku 17 Dravidian Tamil, Telegu 15 Quechuan Quechua (Bolivian), Quechua (Cuzco), Quichua 15 Japanese Japanese 14 Mayan Itzaj Maya, Mam (Southern), Mopan Maya, Tzutujil, Yukateko Maya 14 Altaic Azerbaijani, Turkish 12 Kartvelian Georgian 12 Austro-Asiatic Pnar, Remo 11 Oto-Manguean Mazatec (Jalapa), Zapotec (Yalalog) 10 Burushaski Burushaski 9 Algic Ojibwe 6 Kordofanian Moro 6 Muskogean Chickasaw 6
Language Family Languages Data points Northwest Caucasian Kabardian 6 Pama-Nyungan Warlpiri, Yan-Nhangu 6 Salishan Montana Salish 6 Ticuna Ticuna 6 Uto-Aztecan Paiute (Northern), Ute 6 Wakashan Kwakw'ala 6 Tucanoan Waimaha 5 Eskimo-Aleut Aleut (Eastern), Aleut (Western) 4 Chapacura-Wanham Wari’ 3 Creole Hawaiian Creole 3 Ijoid Defaka 3 Nakh-Dagestanian Udi 3 Tangkic Kayardild 3 Arauan Banawa 2
Relied on primary source descriptions of the laryngeal specifications Aggregate analyses VOT categories Negative: < 0 ms Short-lag: > 0 ms and < 35 ms Long-lag: > 35 ms
voiceless emphatic
voiceless non−emphatic
lax voiced emphatic short−lag
voiceless lax unaspirated
voiceless aspirated
fortis
voiced non−emphatic
voiced unaspirated
plain
voiceless unaspirated
lenis
Kuhl & Miller 1975
category labial coronal dorsal Negative
Short-lag 14 ms 18 ms 30 ms Long-lag 62 ms 65 ms 76 ms
Variation in language-specific VOT means (ms)
Range: -161 to 117 ms Range: -177 to 130 ms Range: -144 to 154 ms Median values
25 50 75 100 125 −180 −95 −10 75 160
labial count
25 50 75 100 125 −180 −95 −10 75 160
coronal count
25 50 75 100 125 −180 −95 −10 75 160
dorsal count
Place differences Canonical order: VOT[labial] < VOT[coronal] < VOT[dorsal]
Comparison Place1 < Place2 Place2 < Place1 N labial - coronal 76% 24% 339 coronal - dorsal 89% 11% 337 labial - dorsal 96% 4% 317
Canonical order Non-canonical order
Maddieson 1997, Cho & Ladefoged 1999, Whalen et al. 2007, Chodroff & Wilson 2017
Aggregate analysis of language-specific VOT means (ms)
http://dev.eleanorchodroff.com/apps/crosslgVOT
r = 0.98* r = 0.97* r = 0.97*
−90 90 180 −180 −90 90 180
coronal dorsal
−90 90 180 −180 −90 90 180
dorsal labial
−90 90 180 −180 −90 90 180
labial coronal
Long-lag VOT
r = 0.83* r = 0.79* r = 0.78*
60 85 110 135 160 35 60 85 110 135 160
coronal dorsal
60 85 110 135 160 35 60 85 110 135 160
dorsal labial
60 85 110 135 160 35 60 85 110 135 160
labial coronal
Short-lag VOT
r = 0.66* r = 0.62* r = 0.46*
20 30 40 10 20 30 40
coronal dorsal
20 30 40 10 20 30 40
dorsal labial
20 30 40 10 20 30 40
labial coronal
Negative VOT
r = 0.95* r = 0.84* r = 0.90*
−150 −120 −90 −60 −30 −180 −135 −90 −45
coronal dorsal
−150 −120 −90 −60 −30 −180 −135 −90 −45
dorsal labial
−150 −120 −90 −60 −30 −180 −135 −90 −45
labial coronal
closed, velar spread Tongue body Glottis DORSAL – continuant … α laryngeal closed, alveolar spread Glottis CORONAL – continuant … α laryngeal Tongue tip closed, labial spread Glottis LABIAL – continuant ... α laryngeal Lips
−90 90 180 −180 −90 90 180
coronal dorsal
−90 90 180 −180 −90 90 180
dorsal labial
−90 90 180 −180 −90 90 180
labial coronal
Mapping from distinctive features to phonetic targets is not independent across segments within a language
place of articulation Phonetic targets for [+s.g.]
[pʰ] [tʰ] [kʰ]
VOT place of articulation
[pʰ] [tʰ] [kʰ] Within the phonetic grammar of a language/talker, the phonetic targets corresponding to a phonological feature value [αF] are (ideally) identical for all segments that are specified [αF]
Applied to long-lag stops: Within a language/speaker, duration and timing of glottal opening gesture relative to stop closure interval should be uniform for all stops specified [+s.g.]
Maddieson 1997, Cho & Ladefoged 1999
Several aerodynamic and biomechanical explanations for VOT variation by place of articulation
Maddieson 1997, Cho & Ladefoged 1999
Claim that differences are automatic presupposes that, for all stops within a laryngeal series, phonetic targets for the laryngeal feature are uniform
Westbury & Keating 1984, Keating 1985
Previous research on VOT: Are place differences in VOT planned or automatic / mechanistic?
Can uniformity be reduced to other known effects and constraints on phonetic realization? Talker physiology / aerodynamics
lag), it is physically possible to produce [ph] with a consistently longer VOT than [kh]
Perceptual dispersion
than would be predicted by dispersion alone
Liljencrants & Lindblom 1972, Schwartz et al. 1997, Flemming 2004
Applies strongly to languages and speakers, thereby ensuring cross-talker relational invariance / restricting individual differences
Each point = pair of VOT means (ms) for a speaker of American English
Chodroff & Wilson 2017
Strong evidence for a uniformity constraint operating on the phonetic implementation of stop consonant laryngeal features Evidence from VOT covariation cross-linguistically Evidence from VOT covariation across talkers of American English Linear relation arises from underlying identity (or near-identity) in the phonetic implementation of laryngeal feature value within each series à Uniform duration and timing of glottal gestures (abduction and adduction) relative to supralaryngeal closure
Role of contrast
à Does uniformity apply as strongly to ‘unpaired’ stops as to those with in minimal laryngeal contrasts (e.g., languages with /p t k/ but /b d/)
Examine cross-linguistic patterns for other features and segments
à Is uniformity specific to stop VOT? Evidence from fricatives in American English and Czech Chodroff 2017 à Do some languages deviate from uniformity (e.g., as the result of recent sound change)?
Relate to phonological theories of feature hierarchies
à Identify natural classes (e.g., stops) strongly bound by uniformity
Andries Coetzee (UMich), James Kirby (Edinburgh), Paul Morris (Iowa), Sharon Rose (UCSD) for generously providing data Doug Whalen (CUNY, Haskins) Paul Smolensky (JHU, Microsoft) Northwestern Phonatics Berkeley Phonetics and Phonology Forum ("Phorum") Northwestern University Postdoctoral Professional Development Travel Grant